Chapman UniversityChapman University Digital Commons

ESI Working Papers Economic Science Institute

2015

Status and the Demand for Visible Goods:Experimental Evidence on ConspicuousConsumptionDavid ClingingsmithCase Western Reserve University, Cleveland, Ohio

Roman M. SheremetaChapman University

Follow this and additional works at: http://digitalcommons.chapman.edu/esi_working_papers

Part of the Econometrics Commons, Economic Theory Commons, and the Other EconomicsCommons

This Article is brought to you for free and open access by the Economic Science Institute at Chapman University Digital Commons. It has beenaccepted for inclusion in ESI Working Papers by an authorized administrator of Chapman University Digital Commons. For more information, pleasecontact laughtin@chapman.edu.

Recommended CitationClingingsmith, D. and Sheremeta, R.M. (2015). Status and the demand for visible goods: Experimental evidence on conspicuousconsumption. ESI Working Paper 15-27. Retrieved from http://digitalcommons.chapman.edu/esi_working_papers/176

Status and the Demand for Visible Goods: Experimental Evidence onConspicuous Consumption

CommentsWorking Paper 15-27

This article is available at Chapman University Digital Commons: http://digitalcommons.chapman.edu/esi_working_papers/176

1

Status and the Demand for Visible Goods:

Experimental Evidence on Conspicuous Consumption

David Clingingsmith a,*

Roman M. Sheremeta a,b

a Weatherhead School of Management, Case Western Reserve University

11119 Bellflower Road, Cleveland, OH 44106, U.S.A. b Economic Science Institute, Chapman University

One University Drive, Orange, CA 92866, U.S.A.

December 4, 2015

Abstract

Some economists argue that consumption of publicly visible goods is driven by

social status. Making a causal inference about this claim is difficult with

observational data. We conduct an experiment in which we vary both whether a

purchase of a physical product is publicly visible or kept private and whether the

income used for purchase is linked to social status or randomly assigned. Making

consumption choices visible leads to a large increase in demand when income is

linked to status, but not otherwise. We investigate the characteristics that mediate

this effect and estimate its impact on welfare.

JEL Classifications: C91, D03

Keywords: status, conspicuous consumption, experiment

* Corresponding author: David Clingingsmith, E-mail: dlc43@case.edu or david.clingingsmith@gmail.com

We thank Yan Chen, Ori Heffetz, David Huffman, John List, Tanya Rosenblat, Klaus Schmidt, Justin Sydnor, Lise

Vesterlund, Alistair Wilson, Bart Wilson and seminar participants at Case Western Reserve University, Chapman

University, Kent State University, the University of Michigan, the University of Pittsburgh as well as participants at

the North American Economic Science Association Meetings in Dallas for helpful comments. We also thank the

Weatherhead School of Management for generous funding of this project and Sarah Mattson for excellent research

assistance. The usual disclaimers apply.

2

1. Introduction

Social status refers to the hierarchical position an individual occupies in society. Social

status is related to an individual’s attributes, such as intelligence, creativity, beauty, affiliation, or

family of origin, either through the returns such attributes earn in economic activity or the esteem

in which they are held by society, or both.1

Veblen argued that social status has a profound influence on a person’s consumption

decisions. His book The Theory of the Leisure Class contends that status concerns affect the

consumption choices of anyone whose income places them above the level of subsistence (2009

[1899]). Social conventions specify a minimum standard of clothing, food, and living conditions

that are acceptable for each status level. Since social status and income are positively correlated,

the acceptable standard of consumption for those of higher status includes more and better goods

than for those of lower status. According to Veblen, an important function of consumption is to

signal high status to others. Consumption choices can only do this to the extent they are both visible

to others and associated with high status. Part of the motivation for wearing a fine suit or driving

a luxury car, both of which are visible to others, is to convey the message to others that one has

high status.

Consumer goods vary in the degree to which the act of consuming them is visible to the

public, and thus in their suitability for serving as public markers of status. Many people see us

when we are in our cars, when we wear our work clothes, or when we eat at a restaurant. Far fewer

1 See a more detailed discussion at http://www.britannica.com/topic/social-status.

3

see our sleepwear, what we have for breakfast at home, or the brand of toilet paper we buy.2 Veblen

termed status signaling via the acquisition and display of visible goods conspicuous consumption.3

A number of studies have presented evidence about the relationship between status and the

consumption of visible goods (Ravina, 2007; Grinblatt et. al. 2008; Charles et. al., 2009; Heffetz,

2011; Kuhn et al., 2011). However, there are several difficulties in identifying conspicuous

consumption as a motivation for consuming visible goods. First, visibility is only one of many

properties possessed by any given good that contribute to the observed demand for it. It is difficult

to disentangle demand for visibility from demand for these other properties. While we may

conjecture that a person buys a Mercedes rather than a Toyota to signal high social status, a

Mercedes is a superior car in many ways besides the signal it sends about status. A second, subtly

related problem is the link between income and social status. Income has effects on consumption

that are independent of any status motivation. Observed correlations between status and

consumption could be pure income effects. Income also tends to be correlated with various

characteristics that confer social status through popular esteem, such as intelligence, education

level, family background, profession, and political clout. Lastly, it is difficult to disentangle

conspicuous consumption from social learning as factors that drive individuals with similar social

status to make similar consumption decisions (Grinblatt et al., 2008).

We tackle the challenge of identifying conspicuous consumption by conducting a

controlled experiment. In our experiment, individuals have an opportunity to purchase a desirable

2 With the rise of social networks and associated technologies such as the smartphone, many formerly private choices,

such as food consumed at home or the decoration of private spaces, have become increasingly visible to the public.

Everyday millions of people post descriptions and images from their private lives on social networks such as Facebook

and Instagram that feature the goods they consume. 3 Several attempts have been made over the years to develop Veblen’s ideas within a more formal microeconomic

framework. Following Leibenstein (1950), some authors have mistakenly attempted to capture Veblen’s argument

with the notion that price is directly a part of utility. Veblen’s analysis implies instead that the determinants of utility

are consumption and social status in the eyes of others. Those goods which signal social status must be visible, but

signaling may occur both through quantity and quality/price (Bagwell and Bernheim, 1996).

4

consumer good: gourmet chocolate truffles. We independently vary both the visibility of

consumption choices to others and whether the income available for consumption is linked to

social status.4 After accruing income, each participant indicates their desired quantity of chocolate

truffles for each of several potential prices. A common price at which sales are actually made is

randomly selected from the potential prices at the end of the experiment.

We manipulate this process in two ways, using a two-by-two design. In the first

manipulation, income is either assigned randomly or based on a participant’s rank. A participant’s

rank is determined how well they do relative to the eleven other participants in their experimental

session on a thirty-minute cognitive test. Our participants are students at Case Western Reserve

University (CWRU). As at many elite universities, cognitive ability confers social status at

CWRU.5 When income is assigned by rank it is directly correlated with status, but when it is

assigned randomly it is unrelated to status. In the second manipulation, communication about the

quantity of truffles purchased is either private, so that only the participant and experimenter know,

or public, so that all participants in the experimental session can see how much each purchased.

Participants know how their choices will be communicated before making them. We refer to the

four treatments as rank-private, rank-public, random-private, and random-public.

Our design allows us to isolate the effect of visibility on demand since all other properties

of the chocolate truffles are identical across the public and private treatments. We can also isolate

the effect of the linkage between income and status. We can rule out social learning as a driver of

4 Our study is related to a large literature on status signaling as a motivation for charitable giving and behavior in

social dilemmas (Andreoni and Petrie, 2004; Soetevent, 2005; Andreoni et al., 2009; Ariely et al., 2009; Bracha and

Vesterlund, 2013; Karlan and McConnell, 2014; Samek and Sheremeta, 2014, 2015). However, our use of a physical

product as a status signal eliminates confounds present in previous studies. Buying chocolate provides only private

benefit to the person who purchased it, while charitable giving and social dilemmas provide benefits to others as well.

Our study therefore does not involve the confounding factors of generosity and altruism present in these other studies. 5 While cognitive ability is a source of social status in general, it is particularly important in the social world of

university students.

5

consumption decisions within the experiment because our participants do not interact and have no

information about the choices of others when they make their decisions.

Veblen’s theory of conspicuous consumption predicts that demand will be higher in rank-

public than rank-private because when status is linked to income, visibility leads people to

consume more to signal status. To the extent that by assigning income randomly we completely

sever the link to status, Veblen’s theory also predicts that demand will be the same in random-

public and random-private.

Consistent with these predictions, we find that making consumption choices publicly

visible leads to a large increase in demand when income is linked to status, but not when income

is assigned randomly. In other words, we find that the necessary conditions for conspicuous

consumption are 1) for income to be correlated with status and 2) for consumption choices to be

publicly visible to others. The effect is quite large: mean quantity demanded is 1.94 truffles in

rank-private and 4.98 truffles in rank-public, an increase of 257%. When income is unrelated to

status, visibility does not induce conspicuous consumption: mean quantity demanded is 1.74 in

random-private and 1.75 in random-public. Although our data provide support for a hypothesis

that status is a significant factor motivating consumption of visible goods, the relationship between

status level and conspicuous consumption is non-monotonic.6 Therefore, a person’s income (and

thus the social status) cannot be easily inferred from their actual chocolate choice in rank-public.

We find that gender and cognitive reflection are important mediators of conspicuous

consumption. Men engaged in conspicuous consumption much more than women. Quantity

6 We find that participants of moderately high and moderately low status engage in conspicuous consumption more

than participants of middle status. Such non-monotonicities are possible in signaling games when players counter-

signal (Spence, 1973; Feltovich et al., 2002). This equilibrium may reflect signaling being relatively cheap: The

lowest-rank person earned $5.25 and the average truffle consumption was 2 in rank-private, so even the lowest-rank

person could inexpensively engage in signaling competition.

6

demanded by men is 477% higher in rank-public than rank-private. Individuals who scored high

on a measure of cognitive reflection (Frederick, 2005), which is the propensity to engage in

conscious deliberation when a situation requires it, also engage more in conspicuous consumption.

We find no impact of risk aversion or competitive social preferences on conspicuous consumption.

Publicly visible choice causes participants to buy chocolate truffles at higher prices than

they would have otherwise. By comparing the demand curves in rank-private and rank-public, we

can estimate both the rank-private consumer surplus and the average welfare loss from making

consumption public. We find that the average welfare loss is as large as the consumer surplus when

the price of chocolate is $0.40, which is approximately the retail price. The negative effect of

conspicuous consumption on welfare loss comes primarily from men, who account for most of the

conspicuous consumption. For women, the loss is much smaller and insignificant.

The non-monotonic relationship between consumption and income means that public

consumption does not convey a credible signal about status, which means that the welfare loss

from conspicuous consumption is not compensated by signaling value of such consumption. To

investigate whether participants derive utility from conspicuous consumption unrelated to

signaling, we asked them to rate their mood at the end of the experiment. For men, we find no

difference in self-reported mood between rank-public and rank-private. To the extent that our

simple mood measure captures the non-consumption externalities of status signaling, it provides

suggestive evidence that the net welfare effect of conspicuous consumption is negative for men.

Women’s moods are low in rank-private, so the net welfare effect of conspicuous consumption

for them is likely positive. However, it is possible that our measure does not fully capture the non-

consumption externalities of status signaling. We discuss some possibilities in the conclusion.

7

We describe the experimental design and procedures in Section 2. Our main results are

presented in Section 3, along with analyses of how conspicuous consumption is related to the level

of status, the characteristics of those who engage in conspicuous consumption, and the welfare

effects of conspicuous consumption. We discuss connections to the literature and implications of

our results in Section 4.

2. Experimental Design and Procedures

The experiment was conducted at Case Western Reserve University. We recruited

participants from an email pool of undergraduate and graduate students. There were 12 sessions

with 12 participants each, for a total of 144 participants. Participants were seated in an ordinary

classroom. The experiment consisted of several parts and participants received instructions

(available in Appendix A) at the beginning of each part.

In all sessions, participants completed a 30-minute cognitive test consisting of 20 multiple-

choice questions. The questions were drawn from a Graduate Record Examination (GRE) test

preparation book (Seltzer, 2009). There were 10 mathematical and 10 verbal questions. All were

of moderate to high difficulty. Participants worked using pencil and paper and recorded their

responses on bubble sheets. The sheets were scanned and scored after 30 minutes had elapsed.

Each participant received one point for each correct response and lost one point for each incorrect

answer. Unanswered questions carried no penalty. Participants were ranked according to the

resulting score, and received a sheet indicating their score and rank among their fellow participants.

We employed the two-by-two design shown in Table 1. The first treatment manipulation

varied the manner in which participants received income. There were 12 income levels between

$5 and $13.25. In the rank treatments, income was allocated based on each participant’s rank on

8

the test. The participant who ranked first on the test got $13.25, while the participant who ranked

last got $5. Participants were given a table showing how rank translated into income with their

own rank and income circled. In the random treatments, income was allocated using a random

draw without replacement. Participants privately drew a card from bag containing cards numbered

1 through 12. Each participant had a table showing how the random numbers translated into

income, from $5 for number 12 to $13.25 for number 1. As a result of this procedure, income and

performance on the test are positively correlated in the rank treatments and uncorrelated in the

random treatments.

Participants were then given an opportunity to spend some of their income on gourmet

chocolate truffles. We chose gourmet chocolate truffles because 1) they are a rival and excludable

consumption good; 2) they are desirable to participants; 3) they are packaged as small, discrete

pieces; and 4) they are of high quality. We wanted a rival and excludable good rather than a public

good, such as a donation to charity, because the benefit of consuming it is purely private. The

motivations underlying demand for rival and excludable goods are less complex, which makes

interpretation of behavior clearer.

Participants completed a table that listed nine potential truffle prices between $0.20 and

$0.60. We asked each participant to indicate how many truffles they would like to purchase at each

of nine potential prices and explained that the roll of a die would later determine the actual price.

They could indicate any quantity between zero and a number exhausting their total income and

would then purchase the indicated quantity corresponding to the actual price. Participants were

told that the remaining cash would be paid to them at the end of the session. Calculators were

provided for this portion.

9

The second treatment manipulation varied how the participants would communicate the

quantity of chocolate they purchased. Before participants completed the table of chocolate choices,

we told them how they would receive their selection once the actual price was determined. In the

private treatments, we explained that we would collect their selection tables and package the

chocolate at the side of the classroom in brown paper bags labeled with their subject numbers. This

would keep everyone’s selections private. Bags would be distributed as participants came up to

get their payments at the end of the experiment. Each participant would get a bag regardless of

whether they purchased any chocolate. In the public treatments, participants were told that after

the actual price was determined, each participant would come up to the whiteboard and write the

quantity of chocolate they selected and the total cost along with their first name and subject

number. We told them that this would speed up our packaging of the chocolate and computation

of payments, which was true.

We collected several other types of data to help us understand what individual

characteristics might mediate the effects of our treatments. Before the GRE test, we measured

cognitive reflection. After participants completed their chocolate choice tables but before the

actual price was determined, we collected measures of risk aversion, social preferences, and

demographic characteristics. The risk aversion and social preferences measures were incentivized.

The three-question cognitive reflection test (CRT) was participants’ first task in the session

(Frederick, 2005). The CRT questions are simple math problems designed to have an intuitively

appealing solution that is incorrect. The test measures an individual’s ability to resist their intuition

and arrive at the correct solution. For example, the first question is “A bat and a ball cost $1.10 in

total. The bat costs $1.00 more than the ball. How much does the ball cost?” The appealing but

incorrect answer is $0.10. The correct answer is $0.05.

10

Participants made a series of 20 binary choices to measure risk aversion (similar to Holt

and Laury, 2002). The choices involved a risk-free amount varying from $0.50 to $10.00 and a

lottery offering a 50% chance to get $10 and a 50% chance to get nothing (see Appendix A). One

of the 20 choices was randomly selected to be paid out at the end of the experiment. We measure

risk aversion using a dummy that identifies whether a participant was more risk averse than the

median.

Next, participants made 12 binary choices to measure social preferences (similar to

Charness and Rabin, 2002). The choices involved additional income for themselves and another

participant with whom they were anonymously paired. Each choice offered the option of $3 to

both self and other or an unequal amount with total value between $3.50 and $8.50 (see Appendix

A). One of the 12 choices was randomly selected to be paid out at the end of the experiment, and

one of the paired participants was randomly selected to be the decision maker, while the other was

selected as a receiver. We use these choices to distinguish participants who always maximize social

welfare from those with competitive preferences. We define a measure of competitive social

preferences as the share of choices in which a participant sacrificed social welfare to increase the

amount by which their payment would be greater than the receiver.

Finally, at the end of the experiment, random draws were conducted to determine payouts

for the risk aversion and social preferences choices as well as the price of chocolate. As the

chocolate was being packaged at the end of the experiment, participants completed a demographic

survey. On average participants earned $15.18 and the experiment lasted for about 70 minutes.

11

3. Results

In this section we present the main results. We start by describing participant characteristics

and our main findings. We then examine how conspicuous consumption is related to status level

and what types of participants are most likely to engage in conspicuous consumption. Finally, we

analyze the welfare effects of conspicuous consumption.

3.1 Participant Characteristics

Table 2 shows the characteristics of our 144 participants. Over three-quarters of our

participants are undergraduate students. They come from a wide range of majors and departments.

Only 13% of participants study economics, finance, or another business-related field. On average,

participants are 20 years old. Gender composition is balanced, with 48% female and 52% male

students. Whites make up 58%, Asians 25%, and African-Americans 8%. The average income in

their family of origin is $141,000. About 42% work in addition to studying, and of those who do,

the average work week is 11.5 hours. Overall, our participant pull is representative of the student

body of Case Western Reserve University.

3.2. Main Findings

Our main question of interest is whether participants engage in conspicuous consumption.

If there is no conspicuous consumption, we should observe no difference in demand for chocolate

across our treatments. If there is conspicuous consumption when income is linked to social status,

we should observe greater demand in rank-public than rank-private. If conspicuous consumption

is about signaling one’s income level itself, we should observe greater demand in random-public

than random-private.

12

Figure 1 shows aggregate chocolate demand curves for each of the four treatments. The

markers show the total quantity demanded at each potential price for the 36 participants in each

treatment. Across all treatments, as standard microeconomic theory predicts, the quantity

demanded falls as price increases. Demand for chocolate in the rank-public treatment is much

higher than the other three treatments. The demand shift is so large that this curve has limited

overlap with the others despite price varying by a factor of three. The curves tell us that making

choices publicly visible increases demand when income is related to status but not when it is

assigned randomly.

We show the average quantity demanded over all potential prices for the four treatments

in Table 3. Mean quantity demanded is 1.94 in rank-private and 4.98 in rank-public, a large and

statistically significant difference of 257%. The total income available to spend on chocolate is the

same in rank-public and rank-private, and its distribution in terms of test performance is also the

same. This allows us to interpret the increased consumption as a causal effect of making

consumption visible. By contrast, when income is assigned randomly rather than by test rank,

mean quantity demanded is 1.74 in random-private and 1.75 in random-public. Again, the total

income available to spend on chocolate is the same in random-public and random-private, and its

distribution in terms of test performance is also the same. This tells us that visibility alone is not

enough to induce an increase in consumption. Income must be correlated with status, here

performance on the test, for public visibility to induce conspicuous consumption. In other words,

the necessary conditions for conspicuous consumption are 1) income must be correlated with status

and 2) consumption choices must be publicly visible to others.

We explore how quantiles of demand vary by treatment for different prices in Table 4.

Panel A compares rank-private and rank-public. All quantiles of demand are higher in the rank-

13

public treatment for every price level. A Wilcoxon rank-sum test at each price level shows the

differences in distributions to be statistically significant. All but one of the p-values are less than

0.05 and most are less than 0.01. Panel B compares random-private and random-public. The

Wilcoxon test shows no statistically significant difference in the distributions at any price other

than $0.25. Overall, the detailed analysis shows that differences found for means from Table 3 are

reflected at all prices and parts of demand curves.

Treatments were randomly assigned to experimental sessions, so the expectation is for

participants in each session to be the same on average in terms of their characteristics. It is

nevertheless possible that participants in different sessions differ in ways important for demand.

To check the robustness of our results, we conduct a regression analysis in which we examine

whether controlling for observable characteristics affects our results.

Let 𝑞𝑖𝑝 be the chocolate demanded by individual 𝑖 when the price is 𝑝, 𝑋𝑖 be a vector of

characteristics for 𝑖, 𝑃𝑢𝑏𝑙𝑖𝑐𝑖 indicate whether consumption choice of 𝑖 is public, and 𝑅𝑎𝑛𝑘𝑖

indicate whether 𝑖’s income was assigned by test rank. Our specification is then

𝑞𝑖𝑝 = 𝛼 + 𝛽𝑃𝑃𝑢𝑏𝑙𝑖𝑐𝑖 + 𝛽𝑅𝑅𝑎𝑛𝑘𝑖 + 𝛽𝑃𝑅(𝑃𝑢𝑏𝑙𝑖𝑐𝑖 × 𝑅𝑎𝑛𝑘𝑖) + 𝑋𝑖′𝜃 + 𝜀𝑖𝑝. (1)

When estimating this regression, we compute standard errors allowing for arbitrary correlation of

𝜀𝑖𝑝 within each individual. Our control vector includes the characteristics we elicited from our

participants as shown in Table 2.

Table 5 reports the results of estimating equation (1) using OLS. Column 1 shows estimates

without any controls. The constant shows average quantity demanded in random-private. Average

quantity demanded in other treatments replicating the means from Table 2 may be obtained by

adding the appropriate coefficients from the set 𝑃𝑢𝑏𝑙𝑖𝑐, 𝑅𝑎𝑛𝑘, and 𝑃𝑢𝑏𝑙𝑖𝑐 × 𝑅𝑎𝑛𝑘. Notably, the

differences in average quantity demanded computed from the regression are 3.04 (p<0.01) between

14

rank-public and rank-private and 0.01 (p=0.99) between random-public and random-private.

Column 2 adds in controls for participant characteristics. Demand is higher for those who have

consumed chocolate in the recent past and lower for those who have eaten in the past five hours,

which makes intuitive sense. Demand is also lower for African-Americans. However, addition of

the controls does not measurably affect the differences between treatments. Conditional on

controls, the differences in average quantity demanded are 3.39 (p<0.01) between rank-public and

rank-private and -0.72 (p=0.26) between random-public and random-private.

Result 1: Making consumption choices visible strongly increases demand when income is

linked to status, but not when income is assigned randomly.

3.3. Levels of Status and Conspicuous Consumption

Chocolate demand is higher when consumption is public for participants whose income

was assigned according to rank, suggesting that participants engage in conspicuous consumption

by buying chocolate to signal their status. In this section, we examine whether conspicuous

consumption varies by status level. If high consumption of visible goods serves as a signal of high

status, we might expect the relationship between a person’s level of status and the degree to which

they engaged in conspicuous consumption to be positive. However, if individuals signal

strategically, then it is also possible to obtain non-monotonic relationship between status and

conspicuous consumption, especially if some participants choose to countersignal their status

(Spence, 1973; Feltovich et al., 2002).

In our experiment, status is conferred by one’s rank on the cognitive test. When income is

assigned by rank, status and income are directly correlated. Previously, we established that

conspicuous consumption takes place when income is assigned by rank. To uncover how

15

conspicuous consumption is related to rank, we plot Engel curves for chocolate demand. The Engel

curve shows how the share of income spent on chocolate varies with income. Figure 2 displays

Engel curves for all four treatments of our experiment. On the vertical axis, participants are binned

by six income levels. On the horizontal axis, the share of income spent on chocolate is averaged

across all nine potential prices for each participant. All four curves appear non-monotonic, though

we must be cautious in interpreting the shapes as there are only six observations behind each data

point. As with total demand, the rank-private Engel curve stands out as distinct from the other

treatments.

We focus on the difference between the rank-private and rank-public curves in Figure 3.

The graph shows the mean effect of visibility on demand for six income/rank levels computed

using regression. Confidence intervals for each difference are shown using dotted lines.

Conspicuous consumption is clearly non-monotonic in status. Participants of moderately low and

moderately high status are most affected. Those in the middle are less affected.

Result 2: Status level has a non-monotonic impact on conspicuous consumption, with

participants of moderately high and moderately low status engaging in conspicuous consumption

more than participants of middle status.

3.4. Who Engaged in Conspicuous Consumption?

In this section we explore to what extent conspicuous consumption is mediated by

individual characteristics such as gender, risk aversion, competitive social preferences, and

cognitive reflection. We collected information about these characteristics because we suspected

that there could be similarities between conspicuous consumption and competitive behavior. In

particular, in signaling through consumption it is important to consume more than others. A

16

number of studies have documented that competitive behavior is linked to gender (Niederle and

Vesterlund, 2007), risk preferences (Cason et al., 2010), and competitive social preferences

(Dohmen and Falk, 2011).7 We also suspected that cognitive reflection (Frederick, 2005) could be

important because our treatment manipulations require participants to be sensitive to a social

setting they are in.

We investigate these mediating factors through a regression of the quantity of chocolate

demanded at the price-individual level on a dummy variable 𝑃𝑢𝑏𝑙𝑖𝑐𝑖 for the public treatment, a

dummy variable 𝑀𝑖 that categorizes the mediating factor, the interaction 𝑃𝑢𝑏𝑙𝑖𝑐𝑖 × 𝑀𝑖, and a

vector of additional controls 𝑋𝑖.

𝑞𝑖𝑝 = 𝛼 + 𝛽𝑃𝑃𝑢𝑏𝑙𝑖𝑐𝑖 + 𝛽𝑃𝑀(𝑃𝑢𝑏𝑙𝑖𝑐𝑖 × 𝑀𝑖) + 𝛾𝑀𝑖 + 𝑋𝑖′𝜃 + 𝜀𝑖𝑝. (2)

The coefficient 𝛽𝑃 measures the effect of making choice public for those who have a zero value

for the mediating factor dummy while 𝛽𝑃𝑀 measures the differential effect of public choice on

those who have a value of one for the dummy. We control flexibly for test rank by including a

dummy variable for each of the 12 test ranks in the control vector 𝑋𝑖.

Table 6 reports the estimation results of specification (2). The unconditional effect on

quantity demanded of making consumption public is 3.04 (column 1). The mediators have no

statistically significant effects of their own on quantity demanded when added as controls, and the

coefficients are much smaller than the effect of public choice, which remains unchanged (column

2). Adding the interaction of public with female gender shows that the effect of public choice

comes entirely from men. The effect for men is 5.93 (p<0.01), while for women it is only 0.55 and

is not statistically significantly different from zero (column 3). Similarly, visibility seems to

primarily affect those individuals who have high CRT scores. The effect for high CRT individuals

7 For a review of this literature see Dechenaux et al. (2015).

17

is 5.29 (p<0.01), while it is only 0.61 and not significantly different from zero for low CRT scorers

(column 4). Recall that these regressions control for cognitive test rank. CRT scores are,

unsurprisingly, correlated with cognitive test scores (Spearman’s 𝜌=0.55). The effect measured

here is therefore for that aspect of CRT not correlated with the cognitive test (e.g., impulsivity of

behavior).8 Participants with higher risk aversion are more affected by visibility, though the effect

is not statistically significant (column 5). Having above-median competitive social preferences

reduce the impact of public consumption (column 6). When we include all mediators in the

regression, gender and cognitive reflection remain important mediators (column 7). The

magnitudes are not much changed from the separate regressions. Interestingly, competitive social

preferences are correlated with gender and CRT, which helps explain why the interaction of

competitive social preferences with public consumption is attenuated in the full regression.

Given that gender and cognitive reflection are the most robust mediators, we now compute

the effect of conspicuous consumption on the four CRT-gender groups. We use the following

regression:

𝑞𝑖𝑝 = 𝛼 + 𝛽1𝑃𝑢𝑏𝑙𝑖𝑐𝑖 + 𝛽2(𝑃𝑢𝑏𝑙𝑖𝑐𝑖 × 𝐹𝑒𝑚𝑎𝑙𝑒𝑖) + 𝛽3(𝑃𝑢𝑏𝑙𝑖𝑐𝑖 × 𝐶𝑅𝑇𝑖) + 𝛽4(𝑃𝑢𝑏𝑙𝑖𝑐𝑖 ×

𝐹𝑒𝑚𝑎𝑙𝑒𝑖 × 𝐶𝑅𝑇𝑖) + 𝛾1𝐹𝑒𝑚𝑎𝑙𝑒𝑖 + 𝛾2𝐶𝑅𝑇𝑖 + 𝛾3(𝐹𝑒𝑚𝑎𝑙𝑒𝑖 × 𝐶𝑅𝑇𝑖) + 𝑋𝑖′𝜃 + 𝜀𝑖𝑝. (3)

In the regression 𝐹𝑒𝑚𝑎𝑙𝑒𝑖 is a dummy variable for female gender and 𝐶𝑅𝑇𝑖 is a dummy for high

CRT.

Table 7 shows the estimated coefficients of the interacted specification (3). We use these

coefficients to compute the effect of making choices in public on the consumption decisions of

each of the four subgroups, which we present in Figure 4. The figure clearly shows that public

8 We would expect the conditional CRT effect to be driven by impulsivity because, unlike with the CRT, inducing

intuitive but incorrect response is not the only goal in the design of questions on the cognitive test.

18

consumption has an effect on all male participants, though the effect is greater on high CRT males

than low CRT males. Public consumption has no significant effect on females. The point estimate

of the effect is greater for high CRT females than low CRT females, though neither are statistically

different from zero.

Result 3: Making consumption choices visible has a large effect on men, particularly those

who exhibit high levels of cognitive reflection. It has no significant effect on women.

3.5. Welfare Effects of Conspicuous Consumption

In this section we examine implications of conspicuous consumption for economic welfare.

Figure 1 shows that the demand for chocolate is much higher in the rank-public treatment than in

the rank-private treatment. This means that at any given price, when consumption is publicly

visible participants choose to buy significantly more chocolate than when such consumption is

private. Figure 1 allows us to see what equilibrium prices would have resulted if we had had a

limited quantity of chocolate and prices were determined over all participants. For example, if we

had 100 chocolates to sell, the price would have been $0.29 in rank-private and $0.51 in rank-

public, a difference of 76%. The share of income spent on chocolate would have been 9% and

15%, respectively.9

In the analysis to follow, we distinguish between two types of welfare effects. One type

results from the excess purchases of chocolate. This is what economists usually mean by welfare.

In addition, we must recognize that the experience of revealing one’s consumption to others may

be inherently pleasant or unpleasant. We use the terms decision welfare and experience welfare to

distinguish between the two (see Kahneman et al., 1999).

9 The total income available to spend on chocolate across all participants in each treatment was $328.50.

19

If we take demand in rank-private as representative of participants’ underlying preferences

for chocolate consumption, the excess consumption of chocolate when choices are public creates

a loss of decision welfare. We can compute both consumer surplus in rank-private and the welfare

loss from the data. Figure 5 shows the basic idea of our approach. First, we fix a market price of

chocolate. We use the participants’ data on chocolate demand to calculate the equilibrium quantity

demanded in rank-private (demand without status signaling) and in rank-public (demand with

status signaling). The area between the market price and the rank-private demand curve is the

consumer surplus in the rank-private treatment. The area between the equilibrium quantity in the

rank-public treatment, given by the intersection of the market price and the rank-public demand

curve, the rank-private demand curve, and the market price is the welfare loss. The welfare loss

represents the excess amount paid for the extra units of chocolate purchased in rank-public over

what would have been paid for them had they actually been purchased in rank-private.

We measured demand in rank-public and rank-private at $0.05 price intervals between

$0.20 and $0.60. To compute the consumer surplus and the welfare loss, we must use this data to

approximate a continuous demand curve. The consumer surplus calculation for a fixed market

price requires an approximation of the rank-private demand curve for all prices for which demand

would be positive. The welfare loss calculation for a fixed market price requires that the rank-

private demand curve be defined for the equilibrium rank-public quantity at that price. We

approximate a continuous curve by fitting a fractional polynomial regression of quantity on price

to the data in each treatment. A fractional polynomial regression allows for a more flexible range

of curve shapes than a standard polynomial by including logarithms, negative powers, and non-

integer powers of the independent variable (Royston and Altman, 1994). We use the subset of

20

powers from the set {x-2, x-1, x-1/2, ln(x), x1/2, x1, x2, x3} that maximizes the likelihood of the model

to construct the curves.

We show the mean demand data and fitted values from the fractional polynomial models

in Figure 6. Circles and diamonds represent, respectively, the mean quantity demanded for rank-

private and rank-public at each elicited price. The solid lines show the fitted curves. The curve for

rank-private shows the out-of-sample extrapolations needed to calculate consumer surplus and the

welfare loss for market prices between $0.20 and $0.60. We do not need to make extrapolations

for rank-public. The rank-private fit uses two terms and the rank-public fit uses four terms. As

Figure 6 shows, the fitted curves match the data quite closely. The R2 for both fractional

polynomial fits exceed 0.99.

Recall that men are much more affected by communicating their consumption choices in

public than women. We show demand data and fitted values separately for men and women in

Figure 7. Note in particular the demand curves for men shown in Panel A. There is almost no

overlap in elicited mean demand between the rank-private and rank-public treatments: participants

purchase as much for $0.60 per piece in rank-public as they do for $0.20 per piece in rank-private.

The shallowing of the rank-private fitted curve to the right of the elicited data depends a lot on the

$0.20 data point, and could in reality be steeper than our estimates.

We present our welfare computations in Table 8. We begin with consumer surplus in the

rank-private treatment for all participants (Panel A). For five fixed prices between $0.20 and

$0.60, we calculate the area between the fitted demand curve, the price, and the zero-quantity line

using a rectangular Riemann approximation with a price delta of $0.001. Mean consumer surplus

ranges from $0.01 for a chocolate price of $0.60 (at which mean demand is only 0.4 pieces) to

$0.76 for a chocolate price of $0.20. The consumer surplus calculations are not very sensitive to

21

assumptions because the fractional polynomial fits the data closely and relatively little out-of-

sample extrapolation is used.

We also use a rectangular Riemann approximation to calculate the welfare loss of making

consumption choices public. These rectangles lie between the rank-private demand curve and a

vertical line descending from the intersection of the fixed price and the rank-public demand curve

to the rank-private demand curve as shown in Figure 5. The mean welfare loss associated with

public consumption choice is also shown in Table 8. The estimates for all participants range from

$0.21 to $0.49 depending on which price we fix. Note that the $0.20 and $0.30 estimates rely on

extrapolation of the rank-private demand curve. The estimated welfare losses are 0.7 to 25 times

as large as consumer surplus. One implication of this is that, except at the subsidized prices of less

than $0.40, participants in the rank-public treatment would have been better off in decision welfare

terms if we had never offered them the chance to buy chocolate. We can conclude that conspicuous

consumption had a negative effect on decision welfare.

Next, we break down the consumer surplus and welfare loss calculations by gender. The

estimates for men are shown in Table 8, Panel B. Men are not very price-sensitive in rank-private,

so their consumer surplus is lower than average. It ranges from $0.01 to $0.57. As might be

expected from the large effects of public choice on men’s demand, the welfare losses for men are

quite large, ranging from $0.52 to $1.00. Given the extrapolation used to calculate these figures,

they should be taken as lower bounds. For women, consumer surplus is somewhat greater, ranging

from zero to $0.94 depending on price. The welfare losses from conspicuous consumption are of

course lower, between $0.01 and $0.13, since women are not effected much by making

consumption choice public.

22

We measured experienced welfare by asking participants to rate their overall mood after

the chocolate distribution was completed. They selected an item from a seven-point scale ranging

from “very bad” to “very good”. Experienced welfare is more difficult to measure than decision

welfare because there is no single dimension like money into which behavioral data can be easily

transformed. It is important to note that, for this reason, our measure does not capture all of the

aspects or dimensions of experience that might bear on an understanding of welfare.

We asked participants about their mood after a number of sources of uncertainty had been

resolved, in particular the actual price of chocolate and payouts for the risk aversion and social

preferences measures.10 Participants had also turned in the sheets with their chocolate choices (in

private treatments) or written their chocolate choices on the white board (in public treatments).

Since along with the treatments, the actual price of chocolate, as well as the risk aversion and

social preferences decisions selected for payout varied at the session level, we want to control for

them when looking at how mood varies by treatments. It turns out that mood is positively

correlated with take-home payouts and, perhaps counterintuitively, the realized price of chocolate

(see regressions reported in Table B1 in Appendix B). In the analysis of how participant mood

varies by treatment, we control for the actual price of chocolate, and payouts from risk aversion

and social preferences elicitation.

We show the regression-adjusted mean mood of participants by treatment in Table 9.11

There is no sizable or statistically significant difference in mood between participants in random-

private and random-public (Panel A). Recall there was also no difference in demand. The average

10 This was done for two reasons. First, we tried to keep each session under 90 minutes and, given that our experiment

was hand run, we were able to substantially shorten each session by having participants answer the survey

questionnaire while their payoffs were calculated. Second, for a more meaningful measure of the experienced welfare

it was important for participants to experience the actual process of status signaling through purchasing chocolate in

the public treatments. 11 The unadjusted mean mood looks very similar and is shown in Table B2 in Appendix B.

23

mood for both of these groups is just over 5, which corresponds to the response “a little bit good.”

Mood does differ between the treatments in which income is assigned by rank. In rank-private,

the average mood is lower at 4.58. But the average mood in rank-public is 5.14, similar to the

random treatments. The difference of 0.56 points is statistically significant at 8%. The effect of

making consumption public has a positive effect on experienced welfare. It is also worth noting

that when consumption is private, assigning income by rank rather than at random has a negative

effect on overall mood.

We break down participant mood by gender in Panels B and C. Recall that male demand

was most strongly affected by public choice. Interestingly, while the mean experienced welfare

for men of 4.94 in rank-private is lower than 5.28 in rank-public, the difference is only -0.34 and

is not statistically significant. The increase in experienced welfare due to public choice comes

entirely through women. For women, mean experienced welfare in rank-private treatment of 3.56

was 1.73 points lower than the 5.29 in the rank public treatment. Even though public consumption

scarcely changed women’s choices, it did make them feel better. Note that experienced welfare in

the random-private and random-public treatments averaged 4.8, which is much closer to rank-

public than rank-private. It might be more correct to say that when income was assigned by rank,

private consumption made women unhappy to a greater extent than public consumption made them

happy.

In summary, the experiment suggests that pubic consumption causes a loss of decision

welfare by inducing more consumption of chocolate. This loss falls primarily on men, who

accounted for most of the conspicuous consumption. The loss exceeds the consumer surplus of

chocolate consumption by a wide margin, so that rank-public men were worse off in terms of

decision welfare than if they had not been offered a chance to buy chocolate. Moreover, for men,

24

we find no differences in self-reported moods between rank-public and rank-private, suggesting

that the net welfare effect of conspicuous consumption is negative for men. For women, we find

that the decision welfare loss is much smaller and insignificant. Also, for women, we find that

conspicuous choice elevated their mood, suggesting an increase in experienced welfare, which

probably results in positive net welfare effect of conspicuous consumption.

Result 4: Conspicuous consumption has a large negative effect on the decision welfare of

men, who accounted for most of the conspicuous consumption. Conspicuous choice elevated the

mood of female participants, but not of males, likely offsetting their small loss in decision welfare.

The net welfare effect of conspicuous consumption is therefore negative for men and positive for

women.

4. Discussion and Conclusion

Standard economic theory suggests that the utility people derive from the consumption of

goods and services drives demand. Veblen proposed that, in addition to consumption utility,

demand for publicly visible goods is also driven by the social signals they send about those who

purchase them.

We use a controlled experiment to examine whether adding the element of visibility to the

purchase of a good induces conspicuous consumption. Our experiment provides clear evidence

that this happens. Although we are not the first to examine social status motivations using an

experiment, our use of a physical good to measure of conspicuous consumption eliminates

confounds present in previous studies. To the best of our knowledge, the only experimental studies

attempting to find evidence of conspicuous consumption use contributions to charities and social

dilemmas. For example, many laboratory and field experiments have found that recognizing

25

donors by revealing their identities increases donations to charities and contributions to public

good (Andreoni and Petrie, 2004; Soetevent, 2005; Ariely et al., 2009; Karlan and McConnell,

2014; Samek and Sheremeta, 2014). One may be tempted to conclude that this change of behavior

is evidence of conspicuous consumption. However, it is not clear whether such change of behavior

is due to social status (Glazer and Konrad, 1996; Hopkins and Kornienko, 2004), or due to other

factors, such as a desire to be seen as generous (Ariely et al., 2009; Benabou and Tirole, 2006;

Andreoni et al., 2009) or to avoid being seen as stingy (Bracha and Vesterlund, 2013; Samek and

Sheremeta, 2014), or perhaps a purely altruistic desire to set an example for others to follow

(Karlan and McConnell, 2014). The nice feature of our experimental design is that, instead of

giving to a charity or another participant, our participants buy chocolate. Since buying chocolate

provides only private benefit to the person who purchased it, we are able to study conspicuous

consumption without confounding factors of generosity and altruism, which are present in other

studies. Further, since participants did not know the consumption decisions of others, our

experiment isolates the effect of conspicuous consumption from social learning (Grinblatt et al.,

2008).

We manipulated not only the visibility of the consumption good but also the link between

status and income. In human societies, income and status tend to be naturally related to one

another. We show that the linkage of income to status is critical for producing the conspicuous

consumption response to the visibility of one’s choices to others. This rules out the hypothesis that

conspicuous consumption is about signaling one’s income level itself. The necessary conditions

for conspicuous consumption are thus that 1) a good be visible to others and that 2) income be an

indicator of one’s status.

26

Although our data provide support for a hypothesis that status is a significant factor

motivating consumption of visible goods, the relationship between status level and conspicuous

consumption is non-monotonic. Specifically, we find that participants of moderately high and

moderately low status engage in conspicuous consumption more than participants of middle status.

This pattern makes it difficult to infer someone’s status from their chocolate purchases. However,

such non-monotonicities are possible in signaling games when players counter-signal (Spence,

1973; Feltovich et al., 2002).12

In addition to finding direct evidence for conspicuous consumption, we also investigated

the characteristics that mediate the effect. We found that men are significantly more likely to

engage in conspicuous consumption than women. This finding contributes to an extensive

literature on gender differences in behavior (Croson and Gneezy, 2009). A possible explanation

why men engage in conspicuous consumption more than women is that men are more competitive

(Niederle and Vesterlund, 2007). As such, when seen by others, men are more likely to express

their competitiveness through generosity (Pan and Houser, 2011), or conspicuous consumption in

our case.13 Alternatively, women may not care about the particular status attribute studied in our

experiment – rank on a cognitive test – as much as men.

We also found that participants who exhibit high levels of cognitive reflection were more

likely to engage in conspicuous consumption. We suspect that participants with high scorers on

12 Another explanation, consistent with the signaling story, is that the low status participants try to avoid being

recognized as the lowest performers on the test and thus purchase more chocolate. Yet another explanation is that

people who performed poorly on the test might use consumption of chocolate to comfort themselves (although we

find no support for this as low status people do not consume much chocolate in other conditions). 13 Another explanation is that women may be discouraged from public consumption of chocolate, since purchasing a

rich and highly caloric treat in public may have more negative connotations for women than men. Indeed, when income

is allocated randomly, women consume less chocolate in the public condition than when consumption is private (see

Table B3 in Appendix B).

27

cognitive reflection test are better at comprehending the experiment and thus are more sensitive to

treatment manipulations.

Finally, we provided estimates of the impact of conspicuous consumption on economic

welfare. Status-seeking behavior may have significant negative effect on economic outcomes by

leading to more aggressive sabotage in work places (Charness et al., 2013) and overly competitive

behavior in contests (Sheremeta, 2015).14 Our study points a more fundamental negative impact

of conspicuous consumption. When people engage in conspicuous consumption, they purchase

more of a good than they would if consumption was private. This leads to a welfare loss for

participants in the experiment 0.7 to 25 times as large as the baseline surplus, depending on prices.

The loss for men, who were most affected by choosing publicly, was 1.7 to 52 times as large as

the baseline surplus. These losses cannot be offset by non-consumption benefits of status signaling

because, in the equilibrium that emerged in our experiment, status cannot be easily inferred from

consumption patterns. Further, we find that the mood of men in our study was no different under

publicly visible or private choice.15

Our findings have practical implications. The fact that we find people engaging in

conspicuous consumption even though there is a little signaling value of such consumption speaks

to the ongoing debate on whether to tax visible goods used for status signaling (Frank, 1999, 2008).

Although such a policy could be warranted, our results contain a puzzle that we think ought to be

resolved before making such a suggestion. Our participants engaged in conspicuous consumption

even though there were no benefits to offset the decision welfare loss. Why did they do so? One

14 Of course, in some environments status-seeking behavior may be beneficial to economy. For example, status and

social recognition may be used to enhance worker performance (Kosfeld and Neckermann, 2011) or to encourage

donations to charities (Karlan and McConnell, 2014; Samek and Sheremeta, 2014, 2015). 15 In a market setting, conspicuous consumption of goods for which public visibility is an integral property will tend

to raise prices. Depending on the shape of the demand curve, this may amplify or mute the decision welfare effect,

though we expect it will be negative.

28

possibility is that they incorrectly expected there to be signaling in equilibrium. If this was the

case, we would expect conspicuous consumption to diminish in repeated trials. It is also possible

that signaling is not the motivation for conspicuous consumption. Sending a signal in our

experiment was fairly cheap, so participants could engage in counter-signaling. However, given a

large difference in income between the highest ranked and the lowest ranked participants, this is

unlikely to be the case.16 It is also possible that there are other non-consumption, non-signaling

benefits to conspicuous consumption that we did not capture in our mood measure.17 We leave

these questions for future research.

16 In our experiment the highest ranked individuals made almost three times more than the lowest ranked individuals.

So, the highest ranked individuals could have easily out-signaled the lowest ranked individuals and still earn more

cash. 17 Sivanathan and Pettit (2010) argue that individuals increase their demand for status goods when they receive

negative information about themselves. They terms such information a self-threat. The response to such a threat is to

engage in conspicuous consumption in order to restore or defend self-integrity. We can view those participants who

ranked in the bottom third on the quiz as having received negative information about themselves. Table B4 in

Appendix B reports estimation of two regressions for participants in the rank-private and rank-public treatments. The

dependent variable in the first regression is the income share spent on consumption and in the second regression the

dependent variable is mood. The independent variables are the dummies indicating treatment, female, having a test

rank > 8, as well as the interactions. The estimation results show that participants who receive negative information

about themselves (i.e., test rank > 8) spend higher share of income on chocolate. Also, we find that there is no effect

on participant mood for this subgroup. These findings are consistent with an interpretation that low-rank participants

have lower mood if they fail to engage in extra chocolate consumption in the rank-public treatment (although the

evidence is not causal).

29

References

Andreoni, J., & Bernheim, B.D. (2009). Social image and the 50–50 norm: A theoretical and

experimental analysis of audience effects. Econometrica, 77, 1607-1636.

Andreoni, J., & Petrie, R. (2004). Public goods experiments without confidentiality: a glimpse into

fund-raising. Journal of Public Economics, 88, 1605-1623.

Ariely, D., Bracha, A., & Meier, S. (2009). Doing good or doing well? Image motivation and

monetary incentives in behaving prosocially. American Economic Review, 544-555.

Bagwell, L.S., & Bernheim, B.D. (1996). Veblen effects in a theory of conspicuous consumption.

American Economic Review, 86, 349-373.

Benabou, R., & Tirole, J. (2006). Incentives and prosocial behavior. American Economic Review,

99, 544-555.

Bracha, A., & Vesterlund, L. (2013). How low can you go? Charity reporting when donations

signal income and generosity. Working Paper.

Cason, T.N., Masters, W.A., & Sheremeta, R.M. (2010). Entry into winner-take-all and

proportional-prize contests: An experimental study. Journal of Public Economics, 94, 604-

611.

Charles, K.K., Hurst, E., & Roussanov, N. (2009). Conspicuous consumption and race. Quarterly

Journal of Economics, 124, 425-67.

Charness, G., & Rabin, M. (2002). Understanding social preferences with simple tests. Quarterly

Journal of Economics, 117, 817-869.

Charness, G., Masclet, D., & Villeval, M.C. (2013). The dark side of competition for status.

Management Science, 60, 38-55.

Croson, R., & Gneezy, U. (2009). Gender differences in preferences. Journal of Economic

Literature, 47, 448-474.

Dechenaux, E., Kovenock, D., & Sheremeta, R.M. (2015). A survey of experimental research on

contests, all-pay auctions and tournaments. Experimental Economics, 18, 609-669.

Dohmen, T. & Falk, A. (2011). Performance pay and multidimensional sorting: Productivity,

preferences, and gender. American Economic Review, 101, 556-90.

Feltovich, N., Harbaugh, R., & To, T. (2002), Too cool for school? Signaling and

countersignalling. RAND Journal of Economics, 33, 630-649.

Frank, R.H. (1999). Luxury fever: Money and happiness in an era of excess. New York: The Free

Press.

Frank, R.H. (2008). Should public policy respond to positional externalities? Journal of Public

Economics, 92, 1777–1786.

Frederick, S. (2005). Cognitive reflection and decision making. Journal of Economic Perspectives,

19, 25-42.

Glazer, A., & Konrad, K. A. (1996). A signaling explanation for charity. American Economic

Review, 86, 1019-1028.

Grinblatt, M., Keloharju, M., & Ikäheimo, S. (2008). Social influence and consumption: Evidence

from the automobile purchases of neighbors. Review of Economics and Statistics, 90, 735-

753.

Heffetz, O. (2011). A test of conspicuous consumption: Visibility and income elasticities. Review

of Economics and Statistics, 93, 1101-1117.

Holt, C.A., & Laury, S.K. (2002). Risk aversion and incentive effects. American Economic Review,

92, 1644-1655.

30

Hopkins, E., & Kornienko, T. (2004). Running to keep in the same place: Consumer choice as a

game of status. American Economic Review, 94, 1085-1107.

Kahneman, D., Diener, E., & Schwarz, N. (1999). Well-being: The foundations of hedonic

psychology. New York: Russell Sage Found.

Karlan, D., & McConnell, M. A. (2014). Hey look at me: The effect of giving circles on giving.

Journal of Economic Behavior and Organization, 106, 402-412.

Kosfeld, M., & Neckermann, S. (2011). Getting more work for nothing? Symbolic awards and

worker performance. American Economic Journal: Microeconomics, 3, 86-99.

Kuhn, P., Kooreman, P., Soetevent, A., & Kapteyn, A. (2011). The effects of lottery prizes on

winners and their neighbors: Evidence from the Dutch postcode lottery. American

Economic Review, 101, 2226–2247.

Leibenstein, H. (1950). Bandwagon, snob, and Veblen effects in the theory of consumers' demand.

Quarterly Journal of Economics, 64, 183-207.

Niederle, M. & Vesterlund, L. (2007). Do women shy away from competition? Do men compete

too much? Quarterly Journal of Economics, 122, 1067-1101.

Pan, X.S., & Houser, D. (2011). Competition for trophies triggers male generosity. PloS One, 6(4),

e18050.

Ravina, E. (2007). Habit formation and keeping up with the Joneses: Evidence from micro data.

Working paper.

Royston, P., & Altman, D.G. (1994). Regression using fractional polynomials of continuous

covariates: Parsimonious parametric modelling. Applied Statistics, 43, 429-467.

Samek, A.S., & Sheremeta, R.M. (2014). Recognizing contributors: An experiment on public

goods. Experimental Economics, 17, 673-690.

Samek, A.S., & Sheremeta, R.M. (2015). Selective recognition: How to recognize donors to

increase charitable giving. Working Paper.

Seltzer, N. (2009). 1,014 GRE practice questions. Princeton Review.

Sheremeta, R.M. (2015). Impulsive behavior in competition: Testing theories of overbidding in

rent-seeking contests. Working Paper.

Sivanathan N., & Pettit, N.C. (2010). Protecting the self through consumption: Status goods as

affirmational commodities. Journal of Experimental Social Psychology, 46, 564‐570.

Soetevent, A. (2005). Anonymity in giving in a natural context – a field experiment in 30 churches.

Journal of Public Economics, 89, 2301-2323.

Spence, M. (1973). Job market signaling. Quarterly Journal of Economics, 87, 355-374.

Veblen, T. (2009 [1899]). The theory of the leisure class: An economic study in the evolution of

institutions. New York: Macmillan

31

Figure 1: Aggregate Demand Curves by Treatment

Notes: Lines plot total quantity demanded for each potential price of chocolate in

each treatment.

32

Figure 2: Engel Curves by Treatment

Notes: Lines plot the quantity demanded averaged across all nine

potential prices for participants in each treatment. Participants are

binned by six income levels, which correspond to test ranks 1 and 2,

3 and 4, etc.

33

Figure 3: Conspicuous Consumption by Income/Status Level

Notes: Graph shows differences in mean income share spent on

chocolate between the rank-private and rank-public treatments for six

income/status bins. Differences and confidence intervals computed

from a regression of mean income share on bin dummies, a public

choice dummy, and their interactions. Confidence intervals allow for

heteroskedasticity.

34

Figure 4: Conspicuous Consumption Effects by Subgroup

Notes: Graph shows estimated difference in mean quantity

demanded for each subgroup between rank-private and rank-public.

Estimates computed from regression that interacts a public treatment

dummy with dummies for female gender and high CRT and their

interaction. Bars show 95% confidence intervals.

35

Figure 5: Welfare Loss from Conspicuous Consumption

36

Figure 6: Fitted Demand Curves for Rank-Public and Rank-Private

Notes: Lines show fitted values from fractional polynomial regressions as

described in Section 3.5.

37

Figure 7: Fitted Demand Curves for Rank-Public and Rank-Private By Gender

Panel A: Men

Panel B: Women

Notes: Lines show fitted values from fractional polynomial regressions as

described in Section 3.5.

38

Table 1: Treatments

Chocolate choice

Income allocation Private Public

Rank 3 sessions,

36 participants

3 sessions,

36 participants

Random 3 sessions,

36 participants

3 sessions,

36 participants

Table 2: Participant Characteristics

Mean SD Min Max

Female 0.48

Age 19.96 1.88 18 23

White 0.58

Asian 0.25

African-American 0.08

Undergraduate 0.78

Study Economics/Business 0.13

Family Income 141K 116K 20K 400K

Employed 0.42

Average Hours if Employed 11.53 10.10 0 40

CRT Score 1.50 1.18 0 3

GRE Test Score 6.80 5.19 -7 14

Risk Aversion 0.54

Competitive Social Preferences 0.14

Had Chocolate Today/Yesterday 0.49

Last Ate < 5 Hours Ago 0.60

39

Table 3: Mean Quantity Demanded by Treatment

Random Rank

H0: Random

= Rank

Private 1.74 1.94 p=0.82

Public 1.75 4.98 p<0.01

H0: Public = Private p=0.99 p<0.01

Notes: Means for each treatment are computed by averaging over all

potential prices. The tests of the null hypotheses are computed using

a pooled regression of quantity demanded on four treatment dummies.

Tests allow for arbitrary correlation of unobservables at the participant

level.

40

Table 4: Distribution of Quantity Demanded for Each Price

Panel A: Income Assignment Based on Rank

Price

(cents)

Pub.

30th

Priv.

30th

Pub.

50th

Priv.

50th

Pub.

70th

Priv.

70th

Pub.

90th

Priv.

90th

Pub. = Priv.

p-value

20 2 0 5 2 12 4 35 10 0.007

25 1 0 4 1 10 3 21 8 0.004

30 1 0 3 0.5 8 2 20 10 0.007

35 1 0 2 0 7 2 13 6 0.007

40 0 0 1 0 6 1 11 5 0.061

45 0 0 1 0 5 1 10 3 0.037

50 0 0 1 0 4 1 9 4 0.035

55 0 0 1 0 3 0 8 3 0.003

60 0 0 0.5 0 3 0 8 1 0.003

Panel B: Income Assignment is Random

Price

(cents)

Pub.

30th

Priv.

30th

Pub.

50th

Priv.

50th

Pub.

70th

Priv.

70th

Pub.

90th

Priv.

90th

Pub. = Priv.

p-value

20 1 0 4 1 6 4 10 10 0.103

25 0 0 3 0 5 2 8 8 0.048

30 0 0 1 0 3 1 5 8 0.176

35 0 0 0 0 2 1 4 8 0.296

40 0 0 0 0 1 1 4 5 0.476

45 0 0 0 0 1 0 2 4 0.310

50 0 0 0 0 1 0 2 3 0.207

55 0 0 0 0 1 0 2 2 0.365

60 0 0 0 0 0 0 2 2 0.337

Notes: P-values are from Wilcoxon tests of the equalities of distributions.

41

Table 5: Effects of Visibility on Demanded

Quantity demanded (1) (2)

Public 0.01 -0.72

(0.69) (0.90)

Rank 0.19 0.10

(0.82) (0.90)

Public × Rank 3.04** 3.49**

(1.29) (1.42)

Female 0.93

(0.74)

High CRT 1.33*

(0.76)

Risk Aversion 0.93

(0.65)

Competitive Preferences -1.19

(1.55)

Ate Chocolate Recently 1.41*

(0.78)

Ate Recently -1.22**

(0.61)

Age 21+ 0.04

(0.84)

White -0.13

(0.80)

African-American -1.96**

(0.93)

Undergraduate 0.51

(0.84)

Family Income > $120K 0.01

(0.67)

Employed 0.42

(0.70)

Constant 1.75*** 0.29

(0.62) (1.68)

Adjusted R2 0.07 0.11

N 1,296 1,296

Notes: * significant at 0.10, ** significant at 0.05, and *** significant

at 0.01 level. Standard errors allow for arbitrary correlation of

unobservables at the participant level.

42

Table 6: Mediators of Conspicuous Consumption

Quantity demanded (1) (2) (3) (4) (5) (6) (7)

Public 3.04*** 3.17*** 5.93*** 0.61 2.03* 4.11*** 2.74

(1.09) (0.87) (1.28) (1.11) (1.09) (1.02) (2.07)

Public × Female -5.38*** -3.79**

(1.58) (1.71)

Public × High CRT 5.29*** 3.93**

(1.71) (1.92)

Public × Risk Aversion 2.41 1.24

(1.87) (1.74)

Public × Competitive Preferences -5.34** -0.67

(2.54) (2.84)

Female -0.21 2.39* -0.96 -0.19 -0.08 1.09

(1.35) (1.41) (1.25) (1.33) (1.33) (1.38)

High CRT 1.41 1.82* -1.37 1.39 1.31 -0.40

(0.97) (0.94) (1.08) (1.00) (0.98) (1.13)

Risk Aversion 0.99 0.84 0.54 -0.16 0.79 -0.07

(0.85) (0.83) (0.85) (1.19) (0.83) (1.17)

Competitive Preferences -1.32 -1.15 -0.50 -1.15 1.81 -0.11

(2.23) (2.15) (2.13) (2.24) (2.75) (2.81)

Constant 1.94*** 2.68 1.91 4.78** 3.21 1.80 3.86

(0.54) (2.58) (2.39) (2.24) (2.73) (2.51) (2.41)

Adjusted R2 0.06 0.23 0.27 0.26 0.23 0.24 0.29

N 648 648 648 648 648 648 648

Notes: * significant at 0.10, ** significant at 0.05, and *** significant at 0.01 level. Regressions of participant

chocolate choices in the rank-private and rank-public treatments. All columns include test rank dummies.

Standard errors allow for arbitrary correlation of unobservables at the participant level.

43

Table 7: Conspicuous Consumption by Gender and CRT

Quantity demanded (1)

Public 3.50**

(1.64)

Public × Female -4.69**

(1.85)

Public × High CRT 3.27

(2.30)

Public × Female × High CRT 1.82

(3.31)

Female × High CRT 1.75

(3.14)

Female 0.01

(1.70)

High CRT -1.61

(1.77)

Constant 4.95**

(2.24)

Adjusted R2 0.29

N 648

Notes: ** significant at 0.05 level. Regressions of

participant chocolate choices in the rank-private and

rank-public treatments. Includes test rank dummies.

Standard errors allow for arbitrary correlation of

unobservables at the participant level.

44

Table 8: Mean Decision Welfare

Panel A: Overall

Chocolate

Price

Rank-Private

Demand

Consumer

Surplus Welfare Loss

Welfare Loss /

Consumer Surplus

$0.60 0.4 $0.01 $0.25 25

$0.50 1.0 $0.08 $0.23 2.9

$0.40 1.8 $0.21 $0.21 1.0

$0.30 2.6 $0.42 $0.35 0.8

$0.20 4.3 $0.76 $0.49 0.7

Panel B: Men

Chocolate

Price

Rank-Private

Demand

Consumer

Surplus Welfare Loss

Welfare Loss /

Consumer Surplus

$0.60 0.4 $0.01 $0.52 52

$0.50 0.7 $0.12 $0.67 5.6

$0.40 1.3 $0.22 $0.71 3.2

$0.30 1.6 $0.36 $0.97 2.7

$0.20 2.9 $0.57 $1.00 1.7

Panel C: Women

Chocolate

Price

Rank-Private

Demand

Consumer

Surplus Welfare Loss

Welfare Loss /

Consumer Surplus

$0.60 0.4 $0.00 $0.13 --

$0.50 0.8 $0.01 $0.04 4.3

$0.40 2.2 $0.17 $0.01 0.05

$0.30 3.7 $0.47 $0.01 0.03

$0.20 5.8 $0.94 $0.07 0.07

Notes: Calculations based on numerical integration of estimated demand curves from the rank-

private and rank-public treatments. See Section 3.5 for an explanation.

45

Table 9: Adjusted Participant Mood by Treatment

Panel A: Overall

Random Rank H0: Rank = Random

Private 5.21 4.38 p=0.05

Public 5.16 5.14 p=0.95

H0: Public = Private p=0.88 p=0.04

Panel B: Men

Random Rank H0: Rank = Random

Private 5.39 5.02 p=0.51

Public 5.91 4.90 p=0.04

H0: Public = Private p=0.16 p=0.79

Panel C: Women

Random Rank H0: Rank = Random

Private 4.95 4.03 p=0.09

Public 3.97 5.47 p=0.01

H0: Public = Private p=0.11 p<0.01

Notes: Mood is measured on a seven point scale ranging from very bad

(1) to very good (7). Adjusted means are calculated from a regression of

mood on treatment dummies and controls for actual chocolate price,

income, and payouts from the risk aversion and social preferences

elicitations. The adjustment assigns each participant the mean level of

income, payouts, and chocolate price.

46

Appendix A – Instructions for the Rank-Public Treatment

PART 1 – EXERCISE

In this exercise you will be asked to answer three questions. Below are three items that vary in

difficulty. Answer as many as you can.

1. A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball.

How much does the ball cost? _____ cents

2. It takes 5 machines 5 minutes to make 5 widgets.

How long does it take 100 machines to make 100 widgets? _____ minutes

3. In a lake, there is a patch of lily pads. Every day, the patch doubles in size. It takes 48 days for

the patch to cover the entire lake.

How long does it take for the patch to cover half of the lake? _____ days

PART 2 – COGNITIVE TEST

You will now take a 30-minute cognitive test containing 20 questions. You may use the margins

of this booklet work out your answer if needed. You may ONLY use pencil, paper, and calculator

provided. No other aids are permitted.

Please use the attached bubble sheet to record your answers. All questions have the following

format:

Who is the current President of the United States?

A. Mitt Romney

B. Bill Clinton

C. Barack Obama

D. George W. Bush

E. David Cameron

To correctly answer this example question, you would fill in bubble C in line 0.

You will gain one point for each correct answer and lose one point for each incorrect answer.

There is no penalty for leaving a question blank. Please try to get as many points as you can.

You will have 30 minutes to work on the questions. You may not be able to finish all the questions

in this time.

COGNITIVE TEST SCORE AND RANK

Your score on the cognitive test is _____

Your rank among today’s test takers is _____/______

This table shows how many dollars you get based on your rank in the cognitive test.

47

Rank Points

1 $13.25

2 $12.50

3 $11.75

4 $11.00

5 $10.25

6 $9.50

7 $8.75

8 $8.00

9 $7.25

10 $6.50

11 $5.75

12 $5.00

PART 3 – CHOCOLATE SELECTION

You have an opportunity to purchase Godiva chocolate truffles using your earnings. Godiva is

considered to be a premium chocolate. The chocolates are individually wrapped. Each weighs

about 1/3 ounce. You will be able to buy as many chocolates as you want as long as the total cost

is less than your earnings.

The price of squares will be between $0.20 and $0.60 per chocolate and will be determined by the

roll of a ten-sided die. You will indicate how many chocolates you would like for each price that

may be randomly selected by the die. In this way you may choose to purchase different amounts

depending on whether the chocolate is relatively expensive or inexpensive.

After you have completed your choices, the actual price will be determined by the roll of a ten-

sided die. Once the actual price is determined, each of you will come up to the board and write out

your first and last name, your ID, the number of chocolates you purchased, the total amount you

spent on chocolate, and your preference for dark or milk chocolate. This will help us package the

chocolate for you.

Use the following table to record your decisions:

Roll of

die

Price How many pieces would you

like to purchase at this price?

Total spent on chocolate at this price =

Price per piece × number of pieces.

1 $0.20 _______ $0.20 × _______ = _______

2 $0.25 _______ $0.25 × _______ = _______

3 $0.30 _______ $0.30 × _______ = _______

4 $0.35 _______ $0.35 × _______ = _______

5 $0.40 _______ $0.40 × _______ = _______

6 $0.45 _______ $0.45 × _______ = _______

7 $0.50 _______ $0.50 × _______ = _______

8 $0.55 _______ $0.55 × _______ = _______

9 $0.60 _______ $0.60 × _______ = _______

48

PART 4 – DECISION PROBLEMS

In this part of the experiment, you will be asked to make a series of choices in decision problems.

How much money you receive will depend partly on chance and partly on the choices you make.

You will see a table with 20 lines. You will state whether you prefer Option A or Option B in each

line. You should think of each line as a separate decision you need to make. At the end of the

experiment, we will draw a card from a deck of cards numbered 1, 2, …., 20. The number on the

card chosen indicates which line in this part will be paid out. You will be paid according to the

option you selected on that line.

In the table, option A always offers a 50% chance to get $10 and a 50% chance to get nothing. To

determine the earnings for people who choose option A, after drawing the card that determines

which line will be paid, we will randomly draw a ball from a bag containing ten orange balls and

ten white balls. That means that when we draw a ball, there is a 50% chance that it is white and a

50% chance that it is orange. If the drawn ball is white and you selected option A in that decision,

you will get the $10, otherwise you will get $0.

Use the following table to record your decisions:

Decision

Number Option A Option B

Choose

A or B

1 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$0.50 for sure

2 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$1.00 for sure

3 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$1.50 for sure

4 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$2.00 for sure

5 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$2.50 for sure 6 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$3.00 for sure

7 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$3.50 for sure

8 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$4.00 for sure

9 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$4.50 for sure

10 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$5.00 for sure 11 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$5.50 for sure

12 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$6.00 for sure

13 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$6.50 for sure

14 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$7.00 for sure

15 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$7.50 for sure

16 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$8.00 for sure

17 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$8.50 for sure

18 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$9.00 for sure

19 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$9.50 for sure

20 $10.00 with 50% chance $0.00 with 50% chance

1,2,3,4,5,6,7,8,9,10,11,

12,13,14,15,16,17,18,1

9,20

$10.00 for sure

49

PART 5 – DECISION PROBLEMS

In this part of the experiment, you will be asked to make a series of choices in decision problems.

You will see a table with 12 lines. You will state whether you prefer Option A or Option B in each

line. You should think of each line as a separate decision you need to make. However, only one

line will be the ‘line that counts’ and will be paid out. In particular, at the end of the experiment,

we will draw a card from a deck of cards numbered 1, 2, …., 12. The number on the card chosen

indicates which line in that part will be paid out.

For each line in the table on the next page, please state whether you prefer option A or option B.

Notice that there are a total of 12 lines in the table – you should think of each line as a separate

decision you need to make.

These decisions affect both your own earnings and those of another participant in the experiment

today. All participants have randomly and anonymously being divided into pairs according to

subject number.

Your earnings for the selected line depend on which option you chose: if you chose option A in

that line, you will receive $3.00 and the other participant who will be matched with you will also

receive $3.00. If you chose option B in that line, you and the other participant will receive earnings

as indicated in the table for that specific line. For example, if you chose B in line 2 and this line is

selected for payment, you will receive $3.00 and the other participant will receive $2.00. Similarly,

if you chose B in line 3 and this line is selected for payment, you will receive $3.00 and the other

participant will receive $1.50.

After you have completed all your choices we will draw a card from a deck of cards numbered 1,

2, …., 12 to determine which line is going to be paid. Within each pair, one participant has the

higher and the other the lower subject number. We will then draw a ball from a bag containing 10

orange and 10 white balls. The ball color determines whether the decision of the higher (orange)

or lower (white) subject number will be implemented. If your decision is chosen to be

implemented, then the earnings to you and the other participant will be determined according to

your choice of A or B. If the other participant’s decision is chosen to be implemented, then the

earnings will be determined according to the other participant choice of A or B.

Use the following table to record your decisions:

Decision

Number

Option A

(you, the other participant)

Option B

(you, the other participant)

Choose

A or B

1 $3.00 to you, $3.00 to other $3.00 to you, $2.50 to other

2 $3.00 to you, $3.00 to other $3.00 to you, $2.00 to other

3 $3.00 to you, $3.00 to other $3.00 to you, $1.50 to other

4 $3.00 to you, $3.00 to other $2.50 to you, $2.00 to other

5 $3.00 to you, $3.00 to other $2.50 to you, $1.50 to other 6 $3.00 to you, $3.00 to other $2.50 to you, $1.00 to other

7 $3.00 to you, $3.00 to other $3.00 to you, $3.50 to other

8 $3.00 to you, $3.00 to other $3.00 to you, $4.00 to other

50

9 $3.00 to you, $3.00 to other $3.00 to you, $4.50 to other

10 $3.00 to you, $3.00 to other $3.50 to you, $4.00 to other 11 $3.00 to you, $3.00 to other $3.50 to you, $4.50 to other

12 $3.00 to you, $3.00 to other $3.50 to you, $5.00 to other

After you have completed your decisions, please fill out the bubble sheet attached to this booklet

with your responses for each line, indicating whether you chose A or B.

51

Appendix B – Additional Tables

Table B1: Participant Mood and Experiment Outcomes

(1) (2)

Income (rank or random) 0.17*** 0.16***

(0.05) (0.05)

Actual Chocolate Price 0.02** 0.02**

(0.01) (0.01)

Payout from Risk Aversion 0.01 0.01

(0.02) (0.02)

Payout from Competitive Preferences 0.18 0.15

(0.21) (0.23)

Female -0.65**

(0.26)

High CRT 0.49*

(0.25)

More Risk Averse 0.17

(0.23)

Competitive Preferences 0.05

(0.54)

Ate Chocolate Recently -0.01

(0.22)

Ate Recently 0.04

(0.25)

Age 21+ 0.00

(0.31)

White -0.27

(0.27)

African-American -0.76*

(0.41)

Undergraduate -0.28

(0.40)

Family Income > $120K -0.08

(0.24)

Employed 0.33

(0.24)

Constant 2.04* 2.60**

(1.06) (1.17)

Adjusted R2 0.11 0.14

N 142 142

Notes: * significant at 0.10, ** significant at 0.05, and ***

significant at 0.01 level. Standard errors corrected for

heteroskedasticity.

52

Table B2: Unadjusted Participant Mood by Treatment

Panel A: Overall

Random Rank H0: Rank = Random

Private 5.11 4.43 p=0.05

Public 5.25 5.12 p=0.68

H0: Public = Private p=0.69 p=0.05

Panel B: Men

Random Rank H0: Rank = Random

Private 5.25 5.28 p=0.95

Public 5.78 4.94 p=0.06

H0: Public = Private p=0.15 p=0.48

Panel C: Women

Random Rank H0: Rank = Random

Private 4.94 3.56 p=0.01

Public 4.65 5.29 p=0.16

H0: Public = Private p=0.60 p<0.01

Notes: Mood is measured on a seven point scale ranging from very bad (1)

to very good (7).

53

Table B3: Mean Quantity Demanded by Treatment and Gender

Panel A: Men

Random Rank

H0: Random

= Rank

Private 0.77 1.36 p=0.41

Public 1.97 6.48 p<0.01

H0: Public = Private p=0.04 p<0.01

Panel B: Women

Random Rank

H0: Random

= Rank

Private 2.97 2.51 p=0.77

Public 1.51 3.46 p=0.08

H0: Public = Private p=0.28 p=0.48

Notes: Means for each treatment are computed by averaging over all

potential prices. The tests of the null hypotheses are computed using a

pooled regression of quantity demanded on four treatment dummies.

Tests allow for arbitrary correlation of unobservables at the participant

level.

54

Table B4: Differential Effects by Test Rank when Income Assigned by Rank

Income Share Spent

on Chocolate

Participant

Mood

(1) (2)

Public 0.11** 0.13

(0.05) (0.65)

Public × Test Rank > 8 0.32* -0.54

(0.18) (1.15)

Public × Female -0.09 1.58**

(0.06) (0.77)

Public × Test Rank > 8 × Female -0.23 1.06

(0.21) (1.52)

Test Rank > 8 -0.01 -0.71

(0.03) (0.98)

Female 0.00 -1.52***

(0.03) (0.49)

Test Rank > 8 × Female 0.12 0.12

(0.08) (1.22)

Constant 0.04** 4.38***

(0.02) (1.08)

Adjusted R2 0.27 0.20

N 648 70

Notes: * significant at 0.10, ** significant at 0.05, and *** significant at

0.01 level. Column (1) is a regression of the share of income spent on

chocolate in the rank-private and rank-public treatments. Column (2) is a

regression of participant mood and includes controls for chocolate price,

and payouts from the risk aversion and social preferences elicitations that

are not shown in the table. Standard errors allow for arbitrary correlation of

unobservables at the participant level.